Antoni Reus schrieb:
||Hi,
Recently I came around a problem testing my app with geronimo 2.1.2
(myfaces 1.2.3).
The app is working well in jboss 4.2 (JSF RI 1.2_04), and I'm trying
it with geronimo 2.1.2 (myfaces 1.2.3).
I have a managed bean called "treeManager", with a "selectedNode"
property that is null the first time.
I have three input components: 2 inputText, and a selectOneMenu, the
JSP code is this:
...
<h:outputLabel for="nodeName" value="Nom"/>
<h:inputText id="nodeName" value="#{treeManager.selectedNode.name}" />
<h:outputLabel for="nodeDescription" value="Descripció"/>
<h:inputText id="nodeDescription"
value="#{treeManager.selectedNode.description}" />
<h:outputLabel for="ambitType" value="Àmbit"/>
<h:selectOneMenu id="ambitType"
value="#{treeManager.selectedNode.ambit}">
<f:selectItem itemLabel="Global" itemValue="global"/>
<f:selectItem itemLabel="Organisme" itemValue="organisme"/>
<f:selectItem itemLabel="Procediment" itemValue="procediment"/>
</h:selectOneMenu>
....
When I try the JSF I get this error.
|javax.faces.FacesException: Exception while calling encodeEnd on
component : {Component-Path : [Class:
org.ajax4jsf.component.AjaxViewRoot,ViewId:
/dissenyador/estructures.jsp][Class:
javax.faces.component.html.HtmlPanelGrid,Id: j_id_jsp_305935947_1]}|
| at
javax.faces.component.UIComponentBase.encodeEnd(UIComponentBase.java:610)|
Caused by: ...
|Caused by: org.apache.jasper.el.JspPropertyNotFoundException:
/dissenyador/estructures.jsp(60,8) '#{treeManager.selectedNode.ambit}'
Target Unreachable, 'selectedNode' returned null|
| at
org.apache.jasper.el.JspValueExpression.getType(JspValueExpression.java:61)|
| at
org.apache.myfaces.shared_impl.renderkit._SharedRendererUtils.findUIOutputConverter(_SharedRendererUtils.java:58)|
| at
org.apache.myfaces.shared_impl.renderkit.RendererUtils.findUIOutputConverter(RendererUtils.java:391)|
| at
org.apache.myfaces.shared_impl.renderkit.html.HtmlRendererUtils.findUIOutputConverterFailSafe(HtmlRendererUtils.java:393)|
| at
org.apache.myfaces.shared_impl.renderkit.html.HtmlRendererUtils.internalRenderSelect(HtmlRendererUtils.java:316)|
| at
org.apache.myfaces.shared_impl.renderkit.html.HtmlRendererUtils.renderMenu(HtmlRendererUtils.java:288)|
| at
org.apache.myfaces.shared_impl.renderkit.html.HtmlMenuRendererBase.encodeEnd(HtmlMenuRendererBase.java:57)|
| at
javax.faces.component.UIComponentBase.encodeEnd(UIComponentBase.java:607)|
Don't know if this is correct, because its true that selectedNode
evaluates to null, but no exception is thrown when rendering the two
previus inputText that also reference selectedNode.
Is this correct? Should I file a bug report?
Interesting. From the stacktrace it look like the problem is when trying
to determine the *type* that this expression returns.
When actually asking for the value, null is simply returned if the
intermediate object is not there.
But when asking what static type of object would be returned from the
bound property, of course there is a real problem if the intermediate
node is not there.
Here's the code from SharedRendererUtils.findUIOutputConverter; the
getType call is the problem:
Class valueType = vb.getType(facesContext); // boom when an
intermediate node in the EL is null
if (valueType == null) return null;
if (String.class.equals(valueType)) return null; //No
converter needed for String type
if (Object.class.equals(valueType)) return null; //There is
no converter for Object class
The HtmlRendererUtils.internalRenderSelect uses
converter = findUIOutputConverterFailSafe
-- which obviously is not quite so "fail safe" :-)
I'm not quite sure what the converter is being used for during rendering
of the select component, but I do know that the rules about converters
and select-components are quite complex. The HTML select component must
always render strings for its options, but JSF requires typed objects to
be passed between the select *component* and the backing beans. So
conversions are required to be invoked at various times.
I think a JIRA issues should definitely be filed for this. If a
converter is optional here, then the code should catch this exception
and not use a converter. Even if a converter is mandatory (ie an error
should be reported if the value-type cannot be determined) then at least
the error reporting needs to be improved. And it is definitely a myfaces
issue, not a Geronimo one.
Just as a side-note, please report issues like this to the user list in
future. The myfaces developers are also subscribed to that...
Regards, Simon
